bam2bcf.c

   1 #include <math.h>
   2 #include <stdint.h>
   3 #include "bam.h"
   4 #include "kstring.h"
   5 #include "bam2bcf.h"
   6 #include "errmod.h"
   7 #include "bcftools/bcf.h"
   8
   9 extern  void ks_introsort_uint32_t(size_t n, uint32_t a[]);
  10
  11 #define CALL_ETA 0.03f
  12 #define CALL_MAX 256
  13 #define CALL_DEFTHETA 0.83f
  14 #define DEF_MAPQ 20
  15
  16 #define CAP_DIST 25
  17
  18 bcf_callaux_t *bcf_call_init(double theta, int min_baseQ)
  19 {
  20         bcf_callaux_t *bca;
  21         if (theta <= 0.) theta = CALL_DEFTHETA;
  22         bca = calloc(1, sizeof(bcf_callaux_t));
  23         bca->capQ = 60;
  24         bca->openQ = 40; bca->extQ = 20; bca->tandemQ = 100;
  25         bca->min_baseQ = min_baseQ;
  26         bca->e = errmod_init(1. - theta);
  27         bca->min_frac = 0.002;
  28         bca->min_support = 1;
  29     bca->per_sample_flt = 0;
  30         return bca;
  31 }
  32
  33 void bcf_call_destroy(bcf_callaux_t *bca)
  34 {
  35         if (bca == 0) return;
  36         errmod_destroy(bca->e);
  37         free(bca->bases); free(bca->inscns); free(bca);
  38 }
  39 /* ref_base is the 4-bit representation of the reference base. It is
  40  * negative if we are looking at an indel. */
  41 int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r)
  42 {
  43     static int *var_pos = NULL, nvar_pos = 0;
  44         int i, n, ref4, is_indel, ori_depth = 0;
  45         memset(r, 0, sizeof(bcf_callret1_t));
  46         if (ref_base >= 0) {
  47                 ref4 = bam_nt16_nt4_table[ref_base];
  48                 is_indel = 0;
  49         } else ref4 = 4, is_indel = 1;
  50         if (_n == 0) return -1;
  51         // enlarge the bases array if necessary
  52         if (bca->max_bases < _n) {
  53                 bca->max_bases = _n;
  54                 kroundup32(bca->max_bases);
  55                 bca->bases = (uint16_t*)realloc(bca->bases, 2 * bca->max_bases);
  56         }
  57         // fill the bases array
  58         for (i = n = r->n_supp = 0; i < _n; ++i) {
  59                 const bam_pileup1_t *p = pl + i;
  60                 int q, b, mapQ, baseQ, is_diff, min_dist, seqQ;
  61                 // set base
  62                 if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
  63                 ++ori_depth;
  64                 baseQ = q = is_indel? p->aux&0xff : (int)bam1_qual(p->b)[p->qpos]; // base/indel quality
  65                 seqQ = is_indel? (p->aux>>8&0xff) : 99;
  66                 if (q < bca->min_baseQ) continue;
  67                 if (q > seqQ) q = seqQ;
  68                 mapQ = p->b->core.qual < 255? p->b->core.qual : DEF_MAPQ; // special case for mapQ==255
  69                 mapQ = mapQ < bca->capQ? mapQ : bca->capQ;
  70                 if (q > mapQ) q = mapQ;
  71                 if (q > 63) q = 63;
  72                 if (q < 4) q = 4;
  73                 if (!is_indel) {
  74                         b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
  75                         b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
  76                         is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
  77                 } else {
  78                         b = p->aux>>16&0x3f;
  79                         is_diff = (b != 0);
  80                 }
  81                 if (is_diff) ++r->n_supp;
  82                 bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
  83                 // collect annotations
  84                 if (b < 4) r->qsum[b] += q;
  85                 ++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
  86                 min_dist = p->b->core.l_qseq - 1 - p->qpos;
  87                 if (min_dist > p->qpos) min_dist = p->qpos;
  88                 if (min_dist > CAP_DIST) min_dist = CAP_DIST;
  89                 r->anno[1<<2|is_diff<<1|0] += baseQ;
  90                 r->anno[1<<2|is_diff<<1|1] += baseQ * baseQ;
  91                 r->anno[2<<2|is_diff<<1|0] += mapQ;
  92                 r->anno[2<<2|is_diff<<1|1] += mapQ * mapQ;
  93                 r->anno[3<<2|is_diff<<1|0] += min_dist;
  94                 r->anno[3<<2|is_diff<<1|1] += min_dist * min_dist;
  95         }
  96         r->depth = n; r->ori_depth = ori_depth;
  97         // glfgen
  98         errmod_cal(bca->e, n, 5, bca->bases, r->p);
  99
 100     // Calculate the Variant Distance Bias (make it optional?)
 101     if ( nvar_pos < _n ) {
 102         nvar_pos = _n;
 103         var_pos = realloc(var_pos,sizeof(int)*nvar_pos);
 104     }
 105     int alt_dp=0, read_len=0;
 106     for (i=0; i<_n; i++) {
 107         const bam_pileup1_t *p = pl + i;
 108         if ( bam1_seqi(bam1_seq(p->b),p->qpos) == ref_base )
 109             continue;
 110
 111         var_pos[alt_dp] = p->qpos;
 112         if ( (bam1_cigar(p->b)[0]&BAM_CIGAR_MASK)==4 )
 113             var_pos[alt_dp] -= bam1_cigar(p->b)[0]>>BAM_CIGAR_SHIFT;
 114
 115         alt_dp++;
 116         read_len += p->b->core.l_qseq;
 117     }
 118     float mvd=0;
 119     int j;
 120     n=0;
 121     for (i=0; i<alt_dp; i++) {
 122         for (j=0; j<i; j++) {
 123             mvd += abs(var_pos[i] - var_pos[j]);
 124             n++;
 125         }
 126     }
 127     r->mvd[0] = n ? mvd/n : 0;
 128     r->mvd[1] = alt_dp;
 129     r->mvd[2] = alt_dp ? read_len/alt_dp : 0;
 130
 131         return r->depth;
 132 }
 133
 134
 135 void calc_vdb(int n, const bcf_callret1_t *calls, bcf_call_t *call)
 136 {
 137     // Variant distance bias. Samples merged by means of DP-weighted average.
 138
 139     float weight=0, tot_prob=0;
 140
 141     int i;
 142     for (i=0; i<n; i++)
 143     {
 144         int mvd      = calls[i].mvd[0];
 145         int dp       = calls[i].mvd[1];
 146         int read_len = calls[i].mvd[2];
 147
 148         if ( dp<2 ) continue;
 149
 150         float prob = 0;
 151         if ( dp==2 )
 152         {
 153             // Exact formula
 154             prob = (mvd==0) ? 1.0/read_len : (read_len-mvd)*2.0/read_len/read_len;
 155         }
 156         else if ( dp==3 )
 157         {
 158             // Sin, quite accurate approximation
 159             float mu = read_len/2.9;
 160             prob = mvd>2*mu ? 0 : sin(mvd*3.14/2/mu) / (4*mu/3.14);
 161         }
 162         else
 163         {
 164             // Scaled gaussian curve, crude approximation, but behaves well. Using fixed depth for bigger depths.
 165             if ( dp>5 )
 166                 dp = 5;
 167             float sigma2 = (read_len/1.9/(dp+1)) * (read_len/1.9/(dp+1));
 168             float norm   = 1.125*sqrt(2*3.14*sigma2);
 169             float mu     = read_len/2.9;
 170             if ( mvd < mu )
 171                 prob = exp(-(mvd-mu)*(mvd-mu)/2/sigma2)/norm;
 172             else
 173                 prob = exp(-(mvd-mu)*(mvd-mu)/3.125/sigma2)/norm;
 174         }
 175
 176         //fprintf(stderr,"dp=%d mvd=%d read_len=%d -> prob=%f\n", dp,mvd,read_len,prob);
 177         tot_prob += prob*dp;
 178         weight += dp;
 179     }
 180     tot_prob = weight ? tot_prob/weight : 1;
 181     //fprintf(stderr,"prob=%f\n", tot_prob);
 182     call->vdb = tot_prob;
 183 }
 184
 185 int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/, bcf_call_t *call)
 186 {
 187         int ref4, i, j, qsum[4];
 188         int64_t tmp;
 189         if (ref_base >= 0) {
 190                 call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
 191                 if (ref4 > 4) ref4 = 4;
 192         } else call->ori_ref = -1, ref4 = 0;
 193         // calculate qsum
 194         memset(qsum, 0, 4 * sizeof(int));
 195         for (i = 0; i < n; ++i)
 196                 for (j = 0; j < 4; ++j)
 197                         qsum[j] += calls[i].qsum[j];
 198         for (j = 0; j < 4; ++j) qsum[j] = qsum[j] << 2 | j;
 199         // find the top 2 alleles
 200         for (i = 1; i < 4; ++i) // insertion sort
 201                 for (j = i; j > 0 && qsum[j] < qsum[j-1]; --j)
 202                         tmp = qsum[j], qsum[j] = qsum[j-1], qsum[j-1] = tmp;
 203         // set the reference allele and alternative allele(s)
 204         for (i = 0; i < 5; ++i) call->a[i] = -1;
 205         call->unseen = -1;
 206         call->a[0] = ref4;
 207         for (i = 3, j = 1; i >= 0; --i) {
 208                 if ((qsum[i]&3) != ref4) {
 209                         if (qsum[i]>>2 != 0) call->a[j++] = qsum[i]&3;
 210                         else break;
 211                 }
 212         }
 213         if (ref_base >= 0) { // for SNPs, find the "unseen" base
 214                 if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
 215                         call->unseen = j, call->a[j++] = qsum[i]&3;
 216                 call->n_alleles = j;
 217         } else {
 218                 call->n_alleles = j;
 219                 if (call->n_alleles == 1) return -1; // no reliable supporting read. stop doing anything
 220         }
 221         // set the PL array
 222         if (call->n < n) {
 223                 call->n = n;
 224                 call->PL = realloc(call->PL, 15 * n);
 225         }
 226         {
 227                 int x, g[15], z;
 228                 double sum_min = 0.;
 229                 x = call->n_alleles * (call->n_alleles + 1) / 2;
 230                 // get the possible genotypes
 231                 for (i = z = 0; i < call->n_alleles; ++i)
 232                         for (j = 0; j <= i; ++j)
 233                                 g[z++] = call->a[j] * 5 + call->a[i];
 234                 for (i = 0; i < n; ++i) {
 235                         uint8_t *PL = call->PL + x * i;
 236                         const bcf_callret1_t *r = calls + i;
 237                         float min = 1e37;
 238                         for (j = 0; j < x; ++j)
 239                                 if (min > r->p[g[j]]) min = r->p[g[j]];
 240                         sum_min += min;
 241                         for (j = 0; j < x; ++j) {
 242                                 int y;
 243                                 y = (int)(r->p[g[j]] - min + .499);
 244                                 if (y > 255) y = 255;
 245                                 PL[j] = y;
 246                         }
 247                 }
 248 //              if (ref_base < 0) fprintf(stderr, "%d,%d,%f,%d\n", call->n_alleles, x, sum_min, call->unseen);
 249                 call->shift = (int)(sum_min + .499);
 250         }
 251         // combine annotations
 252         memset(call->anno, 0, 16 * sizeof(int));
 253         for (i = call->depth = call->ori_depth = 0, tmp = 0; i < n; ++i) {
 254                 call->depth += calls[i].depth;
 255                 call->ori_depth += calls[i].ori_depth;
 256                 for (j = 0; j < 16; ++j) call->anno[j] += calls[i].anno[j];
 257         }
 258
 259     calc_vdb(n, calls, call);
 260
 261         return 0;
 262 }
 263
 264 int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int fmt_flag,
 265                                  const bcf_callaux_t *bca, const char *ref)
 266 {
 267         extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
 268         kstring_t s;
 269         int i, j;
 270         b->n_smpl = bc->n;
 271         b->tid = tid; b->pos = pos; b->qual = 0;
 272         s.s = b->str; s.m = b->m_str; s.l = 0;
 273         kputc('\0', &s);
 274         if (bc->ori_ref < 0) { // an indel
 275                 // write REF
 276                 kputc(ref[pos], &s);
 277                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 278                 kputc('\0', &s);
 279                 // write ALT
 280                 kputc(ref[pos], &s);
 281                 for (i = 1; i < 4; ++i) {
 282                         if (bc->a[i] < 0) break;
 283                         if (i > 1) {
 284                                 kputc(',', &s); kputc(ref[pos], &s);
 285                         }
 286                         if (bca->indel_types[bc->a[i]] < 0) { // deletion
 287                                 for (j = -bca->indel_types[bc->a[i]]; j < bca->indelreg; ++j)
 288                                         kputc(ref[pos+1+j], &s);
 289                         } else { // insertion; cannot be a reference unless a bug
 290                                 char *inscns = &bca->inscns[bc->a[i] * bca->maxins];
 291                                 for (j = 0; j < bca->indel_types[bc->a[i]]; ++j)
 292                                         kputc("ACGTN"[(int)inscns[j]], &s);
 293                                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 294                         }
 295                 }
 296                 kputc('\0', &s);
 297         } else { // a SNP
 298                 kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
 299                 for (i = 1; i < 5; ++i) {
 300                         if (bc->a[i] < 0) break;
 301                         if (i > 1) kputc(',', &s);
 302                         kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
 303                 }
 304                 kputc('\0', &s);
 305         }
 306         kputc('\0', &s);
 307         // INFO
 308         if (bc->ori_ref < 0) kputs("INDEL;", &s);
 309         kputs("DP=", &s); kputw(bc->ori_depth, &s); kputs(";I16=", &s);
 310         for (i = 0; i < 16; ++i) {
 311                 if (i) kputc(',', &s);
 312                 kputw(bc->anno[i], &s);
 313         }
 314     if (bc->vdb != 1)
 315         ksprintf(&s, ";VDB=%.4f", bc->vdb);
 316         kputc('\0', &s);
 317         // FMT
 318         kputs("PL", &s);
 319         if (bcr && fmt_flag) {
 320                 if (fmt_flag & B2B_FMT_DP) kputs(":DP", &s);
 321                 if (fmt_flag & B2B_FMT_DV) kputs(":DV", &s);
 322                 if (fmt_flag & B2B_FMT_SP) kputs(":SP", &s);
 323         }
 324         kputc('\0', &s);
 325         b->m_str = s.m; b->str = s.s; b->l_str = s.l;
 326         bcf_sync(b);
 327         memcpy(b->gi[0].data, bc->PL, b->gi[0].len * bc->n);
 328         if (bcr && fmt_flag) {
 329                 uint16_t *dp = (fmt_flag & B2B_FMT_DP)? b->gi[1].data : 0;
 330                 uint16_t *dv = (fmt_flag & B2B_FMT_DV)? b->gi[1 + ((fmt_flag & B2B_FMT_DP) != 0)].data : 0;
 331                 int32_t  *sp = (fmt_flag & B2B_FMT_SP)? b->gi[1 + ((fmt_flag & B2B_FMT_DP) != 0) + ((fmt_flag & B2B_FMT_DV) != 0)].data : 0;
 332                 for (i = 0; i < bc->n; ++i) {
 333                         bcf_callret1_t *p = bcr + i;
 334                         if (dp) dp[i] = p->depth  < 0xffff? p->depth  : 0xffff;
 335                         if (dv) dv[i] = p->n_supp < 0xffff? p->n_supp : 0xffff;
 336                         if (sp) {
 337                                 if (p->anno[0] + p->anno[1] < 2 || p->anno[2] + p->anno[3] < 2
 338                                         || p->anno[0] + p->anno[2] < 2 || p->anno[1] + p->anno[3] < 2)
 339                                 {
 340                                         sp[i] = 0;
 341                                 } else {
 342                                         double left, right, two;
 343                                         int x;
 344                                         kt_fisher_exact(p->anno[0], p->anno[1], p->anno[2], p->anno[3], &left, &right, &two);
 345                                         x = (int)(-4.343 * log(two) + .499);
 346                                         if (x > 255) x = 255;
 347                                         sp[i] = x;
 348                                 }
 349                         }
 350                 }
 351         }
 352         return 0;
 353 }